Show HN: ClawLite – Local-first personal AI assistant on Telegram

Notifications You must be signed in to change notification settings

Fork 0

Star 1

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

7 Commits

7 Commits

clawlite

clawlite

data

data

docs

docs

installer

installer

tests

tests

.env.example

.env.example

.gitignore

.gitignore

CHANGELOG.md

CHANGELOG.md

CLA.md

CLA.md

CODE_OF_CONDUCT.md

CODE_OF_CONDUCT.md

CONTRIBUTING.md

CONTRIBUTING.md

ClawLite.spec

ClawLite.spec

Dockerfile

Dockerfile

LICENSE

LICENSE

LICENSING_GUIDE.md

LICENSING_GUIDE.md

NOTICE

NOTICE

README.md

README.md

SECURITY.md

SECURITY.md

VERSION

VERSION

docker-compose.yml

docker-compose.yml

requirements-dev.txt

requirements-dev.txt

requirements.txt

requirements.txt

Repository files navigation

Your personal AI assistant. Runs on your machine. Your data stays yours.

Getting Started · Features · Architecture · Contributing

What is ClawLite?

ClawLite is a personal AI assistant that lives in Telegram and runs entirely on your machine.

No subscription. No cloud. No data leaving your device without your explicit permission.

You ask it something — it thinks, searches the web if needed, remembers the context, and answers. Like having a Jarvis, but one that actually respects your privacy.

You: What happened in the Champions League today? ClawLite: [searches web] PSG beat Bayern 5-4 in a thriller at Parc des Princes...

You: Remind me tomorrow to review the match highlights. ClawLite: Got it. I'll remind you tomorrow about the match highlights.

You: What's my name? ClawLite: Fernando. You told me earlier today.

Features

🔒 Local-First by Design Your conversations never leave your machine. ClawLite runs on Ollama — a local LLM that operates with zero outbound traffic. If Ollama is unavailable, it can optionally fall back to Groq cloud, but it tells you explicitly when that happens.

🔍 Real-Time Web Search Powered by Tavily, ClawLite searches the web in real time when it needs current information. Not a cached index — the actual web, right now.

🧠 Persistent Memory ClawLite remembers context across conversations using a local SQLite database with semantic search. It recalls relevant past exchanges without storing anything in the cloud.

🛡️ Sandbox Protection Every tool call and HTTP request is validated against an allowlist before execution. ClawLite cannot access your filesystem, execute arbitrary code, or connect to unauthorized domains.

☁️ Transparent Cloud Fallback If Ollama is unavailable and you have a Groq API key configured, ClawLite uses cloud inference — and always tells you when it does. No silent data transfers.

📰 Daily Brief Enable an optional morning summary: ClawLite searches for relevant news and delivers a personalized brief at the time you choose, every day.

Getting Started

Windows: one-click installer (recommended)

No terminal, no admin rights, no technical knowledge required.

Download the installer from the latest release.

Run ClawLite-Setup-X.Y.Z.exe and follow the wizard.

A setup page opens in your browser asking for your Telegram bot token and Tavily API key — get a bot token from @BotFather on Telegram (/newbot) and a free key from tavily.com, then paste them in.

That's it — ClawLite starts automatically once setup is complete.

Manual install (any OS, developers)

Prerequisites

Python 3.11+

Ollama installed and running

A Telegram account

A Tavily API key (free tier is enough)

1. Get your Telegram bot token

Open Telegram, find @BotFather, and run /newbot. Copy the token it gives you.

1. Clone and configure

git clone https://github.com/forgesynapseltd/ClawLite.git cd ClawLite cp .env.example .env

Edit .env with your credentials:

TELEGRAM_BOT_TOKEN=your_token_here TAVILY_API_KEY=your_tavily_key_here

1. Pull the local model

ollama pull llama3.2

1. Install and run

pip install -r requirements.txt python -m clawlite.main

Or with Docker:

docker compose up -d

Open Telegram, find your bot, and send /start. That's it.

Architecture

ClawLite is built in layers, each with a single responsibility:

Telegram message ↓ Sandbox validation ← blocks malicious content before anything runs ↓ Planner ← decides: search web / recall memory / answer directly ↓ Tool execution ← Tavily search or memory recall ↓ LLM (Ollama local) ← generates the response ↓ Memory store ← saves the exchange to local SQLite ↓ Telegram response

Stack:

python-telegram-bot — Telegram interface

ollama — local LLM inference (llama3.2 by default)

groq — optional cloud fallback

tavily-python — real-time web search

onnxruntime — local embeddings for semantic memory (no PyTorch)

sqlalchemy + SQLite — persistent local storage

loguru — structured logging

Commands

Command Description

/start Introduction and onboarding

/help List of available commands

/status Show current LLM, search, and sandbox status

/memory clear Delete all your conversation history

/brief on Enable daily morning brief

/brief off Disable daily morning brief

Configuration

All configuration lives in .env. Key settings:

Variable Default Description

TELEGRAM_BOT_TOKEN — Required. Get from @BotFather

TAVILY_API_KEY — Required. Get from tavily.com

OLLAMA_MODEL llama3.2 Local model to use

GROQ_API_KEY — Optional. Cloud fallback

SANDBOX_MODE strict strict or permissive

BRIEF_ENABLED false Enable daily brief

BRIEF_HOUR 8 Hour for daily brief (local time)

Why ClawLite?

The AI assistant market has a problem nobody talks about openly.

Every "personal" assistant is personal in name only. Your queries leave your device. Your context is stored on someone else's server. Your usage patterns are being analyzed. And every month, you pay for the privilege.

ClawLite is the alternative. Built on the belief that a truly personal assistant should run on your hardware, search the real web on your behalf, remember your context without that context leaving your machine, and be installable by anyone — not just developers.

No subscription. No cloud dependency. No compromises on what "personal" means.

Roadmap

Telegram bot with local LLM

Real-time web search (Tavily)

Persistent local memory with semantic recall

Sandbox protection

Cloud fallback with transparency

Daily brief scheduler

Voice input (Whisper)

WhatsApp support

Plugin/skill marketplace

One-click desktop installer

Contributing

ClawLite is fully open source under Apache 2.0. Contributions are welcome.

Fork the repository

Create a feature branch: git checkout -b feature/your-feature

Read CLA.md — by submitting a PR you accept it

Submit your pull request

For major changes, open an issue first to discuss what you'd like to change.

License

Apache 2.0 — see LICENSE for details.

ClawLite is a project by ForgeSynapse LTD (Company No. 16692140, England & Wales).

"Your data. Your machine. Your assistant."

⭐ Star this repo if ClawLite is useful to you. It helps more people find it.

About

Your personal AI assistant on Telegram, running entirely on your machine. Open source (Apache 2.0), local-first, no subscription, no cloud lock-in. Real private memory, security by design, and a one-click installer — no terminal, no technical skills needed.

Resources

Readme

License

Apache-2.0 license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy

Uh oh!

There was an error while loading. Please reload this page.

Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks

Report repository

Releases 1

ClawLite v0.4.0 — Automatic Ollama install + Docker step-by-step guide

Latest

Jul 20, 2026

Packages 0

Uh oh!

There was an error while loading. Please reload this page.

Contributors

Uh oh!

There was an error while loading. Please reload this page.

Languages

Python 99.7%

Other 0.3%